2.0.32

• On certificate errors, the server name is now logged instead of the provider name, which is generally more useful.
• IP addresses for DoH servers that require DNS lookups are now cached for at least 12 hours.
• ignore_system_dns is now set to true by default.
• A workaround for a bug in Cisco servers has been implemented.
• A corrupted or incomplete resolvers list is now ignored, keeping the last good known cached list until the next update. In addition, logging was improved and unit tests were also added. Awesome contribution from William Elwood, thanks!
• On Windows, the network probe immediately returned instead of blocking if netprobe_timeout was set to -1. This has been fixed.
• Expired cached IP addresses now have a grace period, to avoid breaking the service if they temporarily can't be refreshed.
• On Windows, the service now returns immediately, solving a long-standing issue when initialization took more than 30 seconds ("The service did not respond to the start or control request in a timely fashion"). Fantastic work by Alison Winters, thanks!
• The SERVER_ERROR error code has been split into two new error codes: NETWORK_ERROR (self-explanatory) and SERVFAIL (a response was returned, but it includes a SERVFAIL error code).
• Responses are now always compressed.